Point prevalence at a specific index date is estimated using contributions to prevalence from both available registry data, and from Monte Carlo simulations of the incidence and survival process, as outlined by Crouch et al (2004) (see References).
prevalence(form, data, num_years_to_estimate, index_date = NULL,
num_reg_years = NULL, cure = 10, N_boot = 1000,
population_size = NULL, proportion = 1e+05, level = 0.95,
population_data = NULL, precision = 2, n_cores = 1, start = NULL)Formula where the LHS is represented by a standard Surv
object, and the RHS has three special function arguments: age, the
column where age is located; sex, the column where sex is located;
entry, the column where dates of entry to the registry are located;
and event, the column where event dates are located.
This formula is used in the following way:
Surv(time, status) ~ age(age_column_name) + sex(sex_column_name) +
entry(entry_column_name) + event(event_column_name)
Using the supplied prevsim dataset, it is therefore called with:
Surv(time, status) ~ age(age) + sex(sex) + entry(entrydate) +
event(eventdate)
A data frame with the corresponding column names provided in
form.
Number of years of data to consider when
estimating point prevalence; multiple values can be specified in a vector.
If any values are greater than the number of years of registry data
available before index_date, incident cases
for the difference will be simulated.
The date at which to estimate point prevalence. Defaults to the latest registry entry date.
The number of years of the registry for which incidence is to be calculated. Defaults to using all available complete years. Note that if more registry years are supplied than the number of years to estimate prevalence for, the survival data from the surplus registry years are still involved in the survival model fitting.
Integer defining cure model assumption for the calculation (in years). A patient who has survived beyond the cure time has a probability of surviving derived from the mortality rate of the general population.
Number of bootstrapped calculations to perform.
Integer corresponding to the size of the population at risk.
The population ratio to estimate prevalence for.
Double representing the desired confidence interval width.
A dataframe that must contain the columns age,
rate, and sex, where each row is the mortality rate for a
person of that age and sex. Ideally, age ranges from [0, 100]. Defaults to
the supplied data; see UKmortality for the format required
for custom datasets.
Integer representing the number of decimal places required.
Number of CPU cores to run the fitting of the bootstrapped
survival models. Defaults to 1; multi-core functionality is provided by the
doParallel package.
Deprecated: Use index_date instead and specify the
number of years of registry data to use with num_reg_years.
Date from which incident cases are included in the format YYYY-MM-DD. Defaults
to the earliest entry date. This value is now inferred by counting back
num_reg_years years of registry data from the index_date. and
An S3 object of class prevalence with the following
attributes:
Estimated prevalence at the index date for
each of the years in num_years_to_estimate.
A list
containing items related to the simulation of prevalence contributions, see
prevalence_simulated
Contributions to
prevalence from each of the supplied registry years, see
prevalence_counted.
The starting date of the registry data included in the estimation.
The index date at which the point prevalence was calculated for.
The known incidence rate for years included in the registry.
Number of years of registry data that were used.
The number of bootstrapped survival models fitted during the calculation.
The p-value resulting from the chi-square test between the simulated and counted prevalent cases for the years of registry data available.
The Surv object used as the response in the survival modeling.
The covariate means from the data.
The most important parameter is num_years_to_estimate, which governs
the number of previous years of data to use when estimating the prevalence at
the index date. If this parameter is greater than the number of years of
known incident cases available in the supplied registry data (specified with
argument num_registry_years), then the remaining
num_years_to_estimate - num_registry_years years of incident data will
be simulated using Monte Carlo simulation.
The larger num_years_to_estimate, the more accurate the prevalence
estimate will be, provided an adequate survival model can be fitted to the
registry data. It is therefore important to provide as much clean registry
data as possible.
Simulated cases are marked with age and sex to enable agreement with population survival data where a cure model is used, and calculation of the posterior distributions of each.
Crouch, Simon, et al. "Determining disease prevalence from incidence and survival using simulation techniques." Cancer epidemiology 38.2 (2014): 193-199.
Other prevalence functions: prevalence_counted,
prevalence_simulated,
test_prevalence_fit
# NOT RUN {
data(prevsim)
# }
# NOT RUN {
prevalence(Surv(time, status) ~ age(age) + sex(sex) + entry(entrydate) + event(eventdate),
data=prevsim, num_years_to_estimate = c(5, 10), population_size=1e6,
index_date = '2013-09-01', num_reg_years = 8,
cure = 5)
prevalence(Surv(time, status) ~ age(age) + sex(sex) + entry(entrydate) + event(eventdate),
data=prevsim, num_years_to_estimate = 5, population_size=1e6)
# Run on multiple cores
prevalence(Surv(time, status) ~ age(age) + sex(sex) + entry(entrydate) + event(eventdate),
data=prevsim, num_years_to_estimate = c(3,5,7), population_size=1e6, n_cores=4)
# }
# NOT RUN {
# }
Run the code above in your browser using DataLab